Classification of Dengue Fever Patients Based on Gene Expression Data Using Support Vector Machines

Background

Symptomatic infection by dengue virus (DENV) can range from dengue fever (DF) to dengue haemorrhagic fever (DHF), however, the determinants of DF or DHF progression are not completely understood. It is hypothesised that host innate immune response factors are involved in modulating the disease outcome and the expression levels of genes involved in this response could be used as early prognostic markers for disease severity.

Methodology/Principal Findings

mRNA expression levels of genes involved in DENV innate immune responses were measured using quantitative real time PCR (qPCR). Here, we present a novel application of the support vector machines (SVM) algorithm to analyze the expression pattern of 12 genes in peripheral blood mononuclear cells (PBMCs) of 28 dengue patients (13 DHF and 15 DF) during acute viral infection. The SVM model was trained using gene expression data of these genes and achieved the highest accuracy of ∼85% with leave-one-out cross-validation. Through selective removal of gene expression data from the SVM model, we have identified seven genes (MYD88, TLR7, TLR3, MDA5, IRF3, IFN-α and CLEC5A) that may be central in differentiating DF patients from DHF, with MYD88 and TLR7 observed to be the most important. Though the individual removal of expression data of five other genes had no impact on the overall accuracy, a significant combined role was observed when the SVM model of the two main genes (MYD88 and TLR7) was re-trained to include the five genes, increasing the overall accuracy to ∼96%.

Conclusions/Significance

Here, we present a novel use of the SVM algorithm to classify DF and DHF patients, as well as to elucidate the significance of the various genes involved. It was observed that seven genes are critical in classifying DF and DHF patients: TLR3, MDA5, IRF3, IFN-α, CLEC5A, and the two most important MYD88 and TLR7. While these preliminary results are promising, further experimental investigation is necessary to validate their specific roles in dengue disease.     

支持向量机及组合预测在蛋白质四级结构分类中的应用

目的:基于支持向量机建立一个自动化识别新肽链四级结构的方法,提高现有方法的识别精度.方法:改进4种已有的蛋白质一级序列特征值提取方法,采用线性和非线性组合预测方法建立一个有效的组合预测模型.结果:以同源二聚体及非同源二聚体为例.对4种特征值提取方法进行改进后其分类精度均提升了2～3%;进一步实施线性与非线性组合预测后,其分类精度再次提高了2～3%,使独立测试集的分类精度达到了90%以上.结论:4种特征值提取方法均较好地反应出蛋白质一级序列包含四级结构信息,组合预测方法能有效地集多种特征值提取方法优势于一体.     

Classification of Nucleotide Sequences Using Support Vector Machines

Species identification is one of the most important issues in biological studies. Due to recent increases in the amount of genomic information available and the development of DNA sequencing technologies, the applicability of using DNA sequences to identify species (commonly referred to as “DNA barcoding”) is being tested in many areas. Several methods have been suggested to identify species using DNA sequences, including similarity scores, analysis of phylogenetic and population genetic information, and detection of species-specific sequence patterns. Although these methods have demonstrated good performance under a range of circumstances, they also have limitations, as they are subject to loss of information, require intensive computation and are sensitive to model mis-specification, and can be difficult to evaluate in terms of the significance of identification. Here, we suggest a new DNA barcoding method in which support vector machine (SVM) procedures are adopted. Our new method is nonparametric and thus is expected to be robust for a wide range of evolutionary scenarios as well as multilocus analyses. Furthermore, we describe bootstrap procedures that can be used to test the significances of species identifications. We implemented a novel conversion technique for transforming sequence data to real-valued vectors, and therefore, bootstrap procedures can be easily combined with our SVM approach. In this study, we present the results of simulation studies and empirical data analyses to demonstrate the performance of our method and discuss its properties.     

Gene function prediction based on Gene Ontology Hierarchy Preserving Hashing

Gene Ontology (GO) uses structured vocabularies (or terms) to describe the molecular functions, biological roles, and cellular locations of gene products in a hierarchical ontology. GO annotations associate genes with GO terms and indicate the given gene products carrying out the biological functions described by the relevant terms. However, predicting correct GO annotations for genes from a massive set of GO terms as defined by GO is a difficult challenge. To combat with this challenge, we introduce a Gene Ontology Hierarchy Preserving Hashing (HPHash) based semantic method for gene function prediction. HPHash firstly measures the taxonomic similarity between GO terms. It then uses a hierarchy preserving hashing technique to keep the hierarchical order between GO terms, and to optimize a series of hashing functions to encode massive GO terms via compact binary codes. After that, HPHash utilizes these hashing functions to project the gene-term association matrix into a low-dimensional one and performs semantic similarity based gene function prediction in the low-dimensional space. Experimental results on three model species (Homo sapiens, Mus musculus and Rattus norvegicus) for interspecies gene function prediction show that HPHash performs better than other related approaches and it is robust to the number of hash functions. In addition, we also take HPHash as a plugin for BLAST based gene function prediction. From the experimental results, HPHash again significantly improves the prediction performance. The codes of HPHash are available at: http://mlda.swu.edu.cn/codes.php?name=HPHash.     

Gene structure conservation aids similarity based gene prediction

One of the primary tasks in deciphering the functional contents of a newly sequenced genome is the identification of its protein coding genes. Existing computational methods for gene prediction include ab initio methods which use the DNA sequence itself as the only source of information, comparative methods using multiple genomic sequences, and similarity based methods which employ the cDNA or protein sequences of related genes to aid the gene prediction. We present here an algorithm implemented in a computer program called Projector which combines comparative and similarity approaches. Projector employs similarity information at the genomic DNA level by directly using known genes annotated on one DNA sequence to predict the corresponding related genes on another DNA sequence. It therefore makes explicit use of the conservation of the exon–intron structure between two related genes in addition to the similarity of their encoded amino acid sequences. We evaluate the performance of Projector by comparing it with the program Genewise on a test set of 491 pairs of independently confirmed mouse and human genes. It is more accurate than Genewise for genes whose proteins are <80% identical, and is suitable for use in a combined gene prediction system where other methods identify well conserved and non-conserved genes, and pseudogenes.     

Gene ontology based transfer learning for protein subcellular localization

Background  

Prediction of protein subcellular localization generally involves many complex factors, and using only one or two aspects of data information may not tell the true story. For this reason, some recent predictive models are deliberately designed to integrate multiple heterogeneous data sources for exploiting multi-aspect protein feature information. Gene ontology, hereinafter referred to as GO, uses a controlled vocabulary to depict biological molecules or gene products in terms of biological process, molecular function and cellular component. With the rapid expansion of annotated protein sequences, gene ontology has become a general protein feature that can be used to construct predictive models in computational biology. Existing models generally either concatenated the GO terms into a flat binary vector or applied majority-vote based ensemble learning for protein subcellular localization, both of which can not estimate the individual discriminative abilities of the three aspects of gene ontology.     

Gene gravity-like algorithm for disease gene prediction based on phenotype-specific network

Background

Polygenic diseases are usually caused by the dysfunction of multiple genes. Unravelling such disease genes is crucial to fully understand the genetic landscape of diseases on molecular level. With the advent of ‘omic’ data era, network-based methods have prominently boosted disease gene discovery. However, how to make better use of different types of data for the prediction of disease genes remains a challenge.

Results

In this study, we improved the performance of disease gene prediction by integrating the similarity of disease phenotype, biological function and network topology. First, for each phenotype, a phenotype-specific network was specially constructed by mapping phenotype similarity information of given phenotype onto the protein-protein interaction (PPI) network. Then, we developed a gene gravity-like algorithm, to score candidate genes based on not only topological similarity but also functional similarity. We tested the proposed network and algorithm by conducting leave-one-out and leave-10%-out cross validation and compared them with state-of-art algorithms. The results showed a preference to phenotype-specific network as well as gene gravity-like algorithm. At last, we tested the predicting capacity of proposed algorithms by test gene set derived from the DisGeNET database. Also, potential disease genes of three polygenic diseases, obesity, prostate cancer and lung cancer, were predicted by proposed methods. We found that the predicted disease genes are highly consistent with literature and database evidence.

Conclusions

The good performance of phenotype-specific networks indicates that phenotype similarity information has positive effect on the prediction of disease genes. The proposed gene gravity-like algorithm outperforms the algorithm of Random Walk with Restart (RWR), implicating its predicting capacity by combing topological similarity with functional similarity. Our work will give an insight to the discovery of disease genes by fusing multiple similarities of genes and diseases.

     

On gene ontology and function annotation

The effort of function annotation does not merely involve associating a gene with some structured vocabulary that describes action. Rather the details of the actions, the components of the actions, the larger context of the actions are important issues that are of direct relevance, because they help understand the biological system to which the gene/protein belongs. Currently Gene Ontology (GO) Consortium offers the most comprehensive sets of relationships to describe gene/protein activity. However, its choice to segregate gene ontology to subdomains of molecular function, biological process and cellular component is creating significant limitations in terms of future scope of use. If we are to understand biology in its total complexity, comprehensive ontologies in larger biological domains are essential. A vigorous discussion on this topic is necessary for the larger benefit of the biological community. I highlight this point because larger-bio-domain ontologies cannot be simply created by integrating subdomain ontologies. Relationships in larger bio-domain-ontologies are more complex due to larger size of the system and are therefore more labor intensive to create. The current limitations of GO will be a handicap in derivation of more complex relationships from the high throughput biology data.     

Contextual weighting for Support Vector Machines in literature mining: an application to gene versus protein name disambiguation

Background  

The ability to distinguish between genes and proteins is essential for understanding biological text. Support Vector Machines (SVMs) have been proven to be very efficient in general data mining tasks. We explore their capability for the gene versus protein name disambiguation task.     

Sparse Support Vector Machines with L_{p} Penalty for Biomarker Identification

The development of high-throughput technology has generated a massive amount of high-dimensional data, and many of them are of discrete type. Robust and efficient learning algorithms such as LASSO [1] are required for feature selection and overfitting control. However, most feature selection algorithms are only applicable to the continuous data type. In this paper, we propose a novel method for sparse support vector machines (SVMs) with L_{p} (p ≪ 1) regularization. Efficient algorithms (LpSVM) are developed for learning the classifier that is applicable to high-dimensional data sets with both discrete and continuous data types. The regularization parameters are estimated through maximizing the area under the ROC curve (AUC) of the cross-validation data. Experimental results on protein sequence and SNP data attest to the accuracy, sparsity, and efficiency of the proposed algorithm. Biomarkers identified with our methods are compared with those from other methods in the literature. The software package in Matlab is available upon request.     

A Scatter-Based Prototype Framework and Multi-Class Extension of Support Vector Machines

We provide a novel interpretation of the dual of support vector machines (SVMs) in terms of scatter with respect to class prototypes and their mean. As a key contribution, we extend this framework to multiple classes, providing a new joint Scatter SVM algorithm, at the level of its binary counterpart in the number of optimization variables. This enables us to implement computationally efficient solvers based on sequential minimal and chunking optimization. As a further contribution, the primal problem formulation is developed in terms of regularized risk minimization and the hinge loss, revealing the score function to be used in the actual classification of test patterns. We investigate Scatter SVM properties related to generalization ability, computational efficiency, sparsity and sensitivity maps, and report promising results.     

Pain Intensity Recognition Rates via Biopotential Feature Patterns with Support Vector Machines

Background

The clinically used methods of pain diagnosis do not allow for objective and robust measurement, and physicians must rely on the patient’s report on the pain sensation. Verbal scales, visual analog scales (VAS) or numeric rating scales (NRS) count among the most common tools, which are restricted to patients with normal mental abilities. There also exist instruments for pain assessment in people with verbal and / or cognitive impairments and instruments for pain assessment in people who are sedated and automated ventilated. However, all these diagnostic methods either have limited reliability and validity or are very time-consuming. In contrast, biopotentials can be automatically analyzed with machine learning algorithms to provide a surrogate measure of pain intensity.

Methods

In this context, we created a database of biopotentials to advance an automated pain recognition system, determine its theoretical testing quality, and optimize its performance. Eighty-five participants were subjected to painful heat stimuli (baseline, pain threshold, two intermediate thresholds, and pain tolerance threshold) under controlled conditions and the signals of electromyography, skin conductance level, and electrocardiography were collected. A total of 159 features were extracted from the mathematical groupings of amplitude, frequency, stationarity, entropy, linearity, variability, and similarity.

Results

We achieved classification rates of 90.94% for baseline vs. pain tolerance threshold and 79.29% for baseline vs. pain threshold. The most selected pain features stemmed from the amplitude and similarity group and were derived from facial electromyography.

Conclusion

The machine learning measurement of pain in patients could provide valuable information for a clinical team and thus support the treatment assessment.     

基于支持向量机的皮肤自体荧光光谱分类方法在糖尿病筛查中的应用

将63例II型糖尿病患者以及140例正常人皮肤的自体荧光光谱分为训练集和测试集两类,针对常用的四种核函数,运用交叉验证、网格寻优法计算最优分类参数,然后结合训练集建模并对测试集分类,结果显示使用径向基核函数时分类效果相对最佳。在此基础上,构建了一种基于线性核函数与径向基核函数的混合核函数,该核函数对人体皮肤自体荧光光谱的分类效果较之于径向基核函数更优,其分类正确率为82．61％,敏感性为69．57％,特异性为95．65％。研究结果表明支持向量机可用于人体皮肤自体荧光光谱的分类,有助于提高糖尿病筛查的正确率。